M.: Probability forecasts for seasonal average anomalies based on GCM ensemble means, COLA

We present two parametric probability methods for forecasting seasonal average anomalies based on the statistical features of the mean of a GCM ensemble. The ensemble-mean probability forecast (EPF) is constructed under the assumption of a normal distribution for the seasonal average ensemble mean. The regressive probability forecast (RPF) is based on regression of the observed seasonal average anomalies on the seasonal average ensemble means. Also discussed, for purpose of comparison, is a commonly used probability forecast scheme based on counting ensemble members in pre-defined categories. This non-parametric scheme will be referred to as the CPF (counting probability forecast). All of these schemes are applied to the nine-member seasonal-averaged ensembles generated using the GCM of the Center for Ocean-Land-Atmosphere Studies. It is pointed out that the standard deviation of the ensemble mean should be the ensemble spread divided by the square root of the ensemble size, rather than the ensemble spread itself. With this standard deviation, the normality assumption of the ensemble mea